»ö« Welcome to Perl 6! | perl6.org/ | evalbot usage: 'p6: say 3;' or rakudo:, std:, or /msg camelia p6: ... | irclog: irc.perl6.org | UTF-8 is our friend!
Set by masak on 12 May 2015.
00:06 mr-foobar left 00:13 Hor|zon joined
japhb Well, there's already ecosystem fallout from TimToady's ~~ change: in particular, Debugger::UI::CommandLine will no longer compile 00:17
00:17 Hor|zon left 00:33 mr-foobar joined
AlexDaniel m: say DateTime.new('1970-01-01T00:00:00').Instant; 00:36
camelia rakudo-moar cf30a7: OUTPUT«Instant:10␤»
AlexDaniel what's up with these 10 seconds?
00:37 cognominal left
b2gills
.oO( how many leap seconds have there been )
 00:43
AlexDaniel b2gills: none? 00:45
b2gills: leap seconds were implemented in 1972, right?
b2gills according to Wikipedia there were 9 from 1972-1979 en.wikipedia.org/wiki/Leap_second 00:49
00:52 llfourn joined
AlexDaniel b2gills: hm, how does that matter? 00:53
I mean, leap seconds should not affect stuff in the past, right? 00:56
dalek ast: 29aa880 | skids++ | integration/error-reporting.t:
Test error and presence of backtrace for return in mainline
 00:59
01:13 BenGoldberg joined 01:18 vendethiel joined 01:19 rmgk_ joined, rmgk left, rmgk_ is now known as rmgk 01:27 rangerprice left 01:29 llfourn left 01:32 tinyblak joined 01:33 tinyblak left 01:34 tinyblak joined 01:43 amurf joined
b2gills Is there any way to bring the following sentence around to talking about multiple dispatch? "Perl 6 like its brethren before it, likes to take ideas from other languages and tries to extend them as far as possible." 02:05
02:13 Hor|zon joined 02:18 Hor|zon left 02:20 bjz joined 02:31 cognominal joined
skids "Perl 6, like its brethren before it, likes to take proven ideas from other languages, and type systems have shown themselves to offer compelling advantages in compiler-side code analysis and powerful user-side semantics via multiple dispatch." 02:32
02:35 isBEKaml_ left 02:38 JimmyZ_ left 02:44 colomon left, colomon joined
japhb Proposal: No pushing incompatible changes *right* after a release. Because then a point release will include the incompatible change. Which happened today. 02:47
02:57 noganex joined
colomon Shouldn’t we be branching from the release tag to do a point release? 02:58
02:59 noganex_ left
b2gills m:m: multi example ( Int $ where 1 ){ say 1 }; multi example ( Int $ where 2 ){ say 2; samewith 1 }; multi example ( Int $ ){ say 3 }; example 2 03:10
m: multi example ( Int $ where 1 ){ say 1 }; multi example ( Int $ where 2 ){ say 2; samewith 1 }; multi example ( Int $ ){ say 3 }; example 2
camelia rakudo-moar cf30a7: OUTPUT«2␤1␤»
b2gills m: multi example ( Int $ where 1 ){ say 1 }; multi example ( Int $ where 2 ){ say 2; nextwith 1 }; multi example ( Int $ ){ say 3 }; example 2
camelia rakudo-moar cf30a7: OUTPUT«2␤3␤»
b2gills skids++: thanks
03:11 ggoebel left 03:24 vendethiel left
b2gills Let's see if my change stays on en.wikipedia.org/wiki/Multiple_dispatch#Perl_6 03:25
skids b2gills++ 03:29
03:30 llfourn joined
b2gills I had to mark the source as "perl" instead of "perl6" to get decent code highlighting 03:30
03:34 itz_ joined 03:35 cognominal left 03:37 itz left, tinyblak left 03:38 tinyblak joined 03:41 yqt left 03:42 tinyblak left 03:55 tinyblak joined 03:56 gfldex left 04:01 tinyblak left, tinyblak joined 04:04 eternaleye left, konsolebox joined 04:05 eternaleye joined 04:06 tinyblak left 04:12 TEttinger left 04:14 Hor|zon joined 04:19 Hor|zon left 04:24 tinyblak joined
ugexe more CLI-janitoring blog posting ugexe.com/create-a-perl6-terminal-progress-bar/ 04:26
04:59 khw left
[Coke] in backscroll, I also would like consistency in naming. At least something that can be explained (this kind uses camelcase. This kind uses kebob. This has underscores. And the fewer the rules, the better.) 05:16
labster m: my $x = 15; $x.sprintf("%04d"); 05:17
camelia rakudo-moar cf30a7: OUTPUT«Directives specify 0 arguments, but 1 argument was supplied␤␤»
smls m: my $x = 15; say $x.fmt("%04d");
camelia rakudo-moar cf30a7: OUTPUT«0015␤»
labster ah, reverse, thanks. 05:18
[Coke] (branching from release tag). yes. post birthday, I hope we pick a git branching strategy that avoids that problem.
labster m: my $x = 15; "%04d".sprintf($x); 05:19
camelia ( no output )
05:20 tinyblak left
smls branching strategy: How about a crazy one like the Qt project, with like 5 layers of branches for different levels of stability/release-proximity, and complicated rules backporting commits upwards/downwards :P 05:21
ShimmerFairy [Coke]: As a semi-random first pick, how's about "Type names are CamelCase (native types camelCase), lower-case function/variable/etc. names are kebab-case, upper-case function/variable/etc. names are UNDERSCORE_CASE" ? 05:22
smls except for that last one, that already seems to be the convention 05:23
05:23 BenGoldberg left
ShimmerFairy And internal/lower-level stuff stays purposefully deviant from whatever rules are picked :P 05:24
smls it's just that not all (older?) names have been brought in line with it
AlexDaniel even if there are lots of complicatd rules, these should be documented somewhere
e
smls and in case of internal / MOP ones, jnthn might be right that it's not worth changing them
ShimmerFairy smls: I like the idea that the more internal stuff has abnormal conventions; it'll make it feel weird if you find yourself using it :) 05:25
.oO(Wait, why am I using is_type and AT-KEY ??)
smls ShimmerFairy: At one point not too long ago, a whole bunch of UNDERSCORE_CASE uppercase methods were changed to KEBAP-CASE 05:27
so I don't think it's realistic to expect them to change in the opposite direction
05:27 mr-foobar left
ShimmerFairy smls: I'm well aware, and I will oppose the idiotic change to YELLING-KEBAB to the death :) 05:27
smls doesn't mind much either way 05:29
ShimmerFairy I do think we need a set naming convention. Even if everyone else agrees on YELLING-KEBAB, at least then I can grumble at the set standard instead of at random unwarranted changes and deprecations :P 05:35
05:36 mr-foobar joined
skids agrees about YELLING_KEBAB/thought the whole reason lc-kebab was popular was because you didn't have to toggle shift. 05:57
erm, yeah see I didn't even manage to toggle shoft there :-)
But what's done is done AFAIC. 05:58
ShimmerFairy skids: yeah, after I got used to - in identifiers, I've basically treated _ as an uppercase - (that's even true, on a US keyboard: Shift+- -> _) 05:59
skids I guess it is immune to capslock though. 06:01
(a key which I should really retask to something even more useless since it is in the way so much) 06:02
ShimmerFairy skids: Yeah, that's true, though since I always use shift and caps lock sits there dead that doesn't apply to me. (I remember one night translating a grammar to P6 grammar that had uppercase rule names, and having a sore pinky from holding shift by the end :P) 06:03
06:05 AlexDaniel left
ShimmerFairy I do find it really interesting how, in programming, _ basically _is_ a letter by now. Makes me wonder what kind of linguistic perspective you can put on the matter... :) 06:05
06:15 domidumont joined, Hor|zon joined 06:19 domidumont left, Woodi left 06:20 Woodi joined, domidumont joined
skids Well it is basically a heirarchal space, on the other sideof space from , ; and … so my perspective is we crashed the linguistic stack. 06:20
06:20 Hor|zon left
Woodi hi #perl6 :) 06:21
ShimmerFairy skids: I'm sure a linguistic study of programming languages would be quite interesting... 06:22
ShimmerFairy quickly glances over at TimToady
skids o/ Woodi 06:23
ShimmerFairy o/
06:26 domidumont left 06:29 jack_rabbit joined
Woodi so what a lists, arrays, hashes, sets, etc. are ? categories, shapes, solids ? 06:31
TimToady: maybe we want some positional in one direction "solid" ? prev positions can be forgetted. such distinction would allow to have simple array/lists implementation 06:34
"streams" ?
skids streams has too much association with characters/bytes. 06:35
Woodi yes, but it's like river, you can't enter into same two times :)
smls .tell masak You offered to discuss parsing binary blobs, right? I wrote down some thoughts on it here: gist.github.com/smls/bc5d0fb42f199574e339 06:37
yoleaux smls: I'll pass your message to masak.
ShimmerFairy smls: I've not yet read past the table of contents, but I thought I'd let you know that I've always thought it'd be better to design a binary parsing thing without worrying about how to cram it into the existing regex language. Not that it can't be done, but ignoring the suggestion that regexes should do it would likely open more possible solutions :) 06:41
smls but also more reinventing of wheels I suspect 06:42
06:42 TEttinger joined
smls although yes, exploring other avenues would be interesting too 06:42
skids Personally I think regex is very capable of the parse side of things, but the Grammar::Generative side of things and the memory management issues make write access a bit dodgy. 06:43
If I were to pursue a different avenue it would be structured OO wih some MOP suger to make the REPRs handle thins like TLV fields and endianness nicely. 06:44
Woodi Perl5 'pack's was nice for geting hunks of binary data, some "templates" could be standarized into \T form :) 06:45
ShimmerFairy I'm not familiar with P5, but from what I've seen "pack" was a nice tool. Even though a lot of it can be done with other existing Perl 6 things, a more readable (and potentially expressive) form of its syntax would be something to consider. 06:46
Woodi it's a bit like fread(3), it takes how much to read and how big each hunk is 06:47
ShimmerFairy The regex modifiers like :bytes are pre-NFG I believe (and _definitely_ pre-S15), and part of the long-overdue changes I should've made ages ago :) 06:48
smls at the very bottom of the gist, there's an example of how the regex-based thing could potentially be used as an 'unpack' alternative
but that just covers 'unpack', not 'pack'
Woodi yes, unpack have *more* sense with reading data :) 06:50
06:51 skids left 06:54 kaare_ joined 06:59 gfldex joined
itz_ twitter++ # image caption "C++ The Good Parts", image K&R 07:03
07:05 mr-foobar left
ShimmerFairy smls: As a note, I don't envision the :bytes modifier existing much longer, as that comes from an earlier conception of how to do the different levels of Unicode in strings. (I expect :bytes to end up just like the Str.bytes method as seen in S32::Str, in other words) 07:07
smls How do you think it will be handled then? 07:08
ShimmerFairy smls: If regexes are still meant to handle binary parsing, I expect the adverb would be called :bin or :binary (like the :bin adverb in IO.open)
07:09 [Tux] joined
smls makes sense 07:09
ShimmerFairy A more general name like :bin would incidentally not restrict you to 8-bit (or more generally 1-byte) units of data, as a bonus :)
07:12 darutoko joined
ShimmerFairy smls: I actually like the rule for handling Pairs. It sounds weird at first but it quickly reminds one of tr/// (if not outright is tr///) 07:14
smls is was somewhat inspired by doc.perl6.org/routine/rotor :) 07:15
But those details are not important; I just tried to develop *one* possible API for it, to bring to light potential edge cases / problems 07:16
ShimmerFairy If I could come with a binary equivalent for the word "grammar" ("grammar" is to "text parsing" as ??? is to "bit/byte/* parsing"), I'd suggest using it as a keyword for the binary version of grammars :P
*come up with 07:18
07:20 RabidGravy joined 07:22 bjz left 07:31 Hor|zon joined 07:36 Hor|zon left
Woodi ShimmerFairy: I was thinking: it could be somename from electronics but only found that that damn they have 40Gb/s network in 40km cables! 07:42
07:43 brrt joined
Woodi over here 2km from "cental" ppls do not pay for internet via cable... 07:43
brrt \o 07:45
07:46 smls left 07:48 bin_005 left 07:49 bin_005 joined
Woodi hi brrt 07:51
brrt fwiw, the algorithm i needed looks like this gist.github.com/anonymous/19228f32eaf1fb4c6802 07:52
i need it to generate all rulesets possible for a given rule head
i.e. all states in the state machine 07:53
RabidGravy morning
brrt morning Woodi, RabidGravy
RabidGravy Rarr!
brrt likes how you can write perl (5) code like it's C or scheme or anything-you-like 07:56
RabidGravy dear c++, I hate you : "no known conversion for argument 1 from ‘const SAMPLETYPE* {aka const short int*}’ to ‘const SAMPLETYPE* {aka const float*}’" 08:01
how? how does it even think that?
ShimmerFairy unions? 08:04
08:04 Averna joined 08:11 brrt left
RabidGravy it's a method with a single argument which is an array the type of which is supposed to be in the SAMPLETYPE define, this is set in the includes 08:11
08:14 kaare__ joined 08:15 kaare_ left
cdc ugexe++ # bloging 08:17
smls++ # bloging (in a gist) 08:22
08:25 telex left 08:26 telex joined 08:27 brrt joined 08:41 vendethiel joined 08:47 brrt left, ShimmerFairy joined, espadrine_ joined 08:54 mr-foobar joined
RabidGravy gave in and manually added the appropriate typedef, which is odd as the include file is supposed to do it 09:01
[ptc] m: class A { has $.a }; say A.new(a => 5) eqv A.new(a => 5); 09:03
botsnack
hi everyone :-)
where have all the bots gone? 09:04
moritz .tell masak the rakudo+nqp 2015.07.1 releases don't work for a star release: "The nqp-m binary is too old" (probably due to the missed version bumb
yoleaux moritz: I'll pass your message to masak.
moritz \o [ptc]; humansnack
09:04 dalek left, dalek joined, ChanServ sets mode: +v dalek
camelia rakudo-moar cf30a7: OUTPUT«False␤» 09:04
synbot6 om nom nom
[ptc] moritz: om nom nom
moritz: thanks! 09:05
does anyone know why class A { has $.a }; say A.new(a => 5) eqv A.new(a => 5); returns False? One would think that the two objects would be equivalent
moritz [ptc]: because the fallback for eqv is ===
[ptc] ... and if there isn't a way to canonically compare the objects, then === is used? 09:06
09:07 bjz joined
moritz well, currently it's just a bunch of multis 09:07
and the Any, Any multi uses ===
[ptc] shouldn't fall back to something along the lines of comparing the equivalence of the .perl of the two objects? 09:08
moritz and there more specialized variants for "interesting" types
[ptc]: well, .perl omits private attributes
m: class B { has $!a; submethod BUILD(:$a!) { } }; say B.new(a => 1).perl eq B.new(a => 2).perl 09:09
camelia rakudo-moar cf30a7: OUTPUT«True␤»
[ptc] is now really confused
why are two objects which have different contents now equal? 09:10
m: class B { has $!a; submethod BUILD(:$a!) { } }; say B.new(a => 1).perl eqv B.new(a => 2).perl
camelia rakudo-moar cf30a7: OUTPUT«True␤»
moritz I never said the objects are equal
just their .perl output
[ptc] ah
moritz and .perl omits private attributes, becuase in general, you can't initialize them by passing them to .new anyway
and to not break encapsulation
so it'd be kinda dangerous to use .perl for eqv semantics 09:11
[ptc] ok. That makes sense
am trying to understand what's going on in order to make the docs for eqv a bit clearer 09:12
09:12 bjz left
moritz other languages have the concept of "hashable" 09:13
and you can only use stuff as hash keys or for eqv comparisons if they are hashable
[ptc] ok. that makes sense too 09:14
moritz I'd very much like Perl 6 to take that route
[ptc] so is there an easy way to explain why two instances of `class A { has $.a };` each instantiated with the same value don't evaluate as being equivalent? I know you said eqv falls back to ===, but I'm not 100% clear about all implications of what that means 09:15
Woodi hash works like separate guid in every object or as hash function result from object ? 09:16
or some attributes equality ? 09:17
ShimmerFairy I forget if there was a better method, but I _think_ you have to implement a new infix:<eqv> multi to have a working eqv (like e.g. multi infix:<eqv>(A $l, A $r) { $l.a eqv $r.a } )
[ptc] or is it simply not possible to compare objects? And that eqv is best thought of when one compares basic data structures such as hashes, lists etc?
09:17 bjz joined
ShimmerFairy "working" as in "not a synonym for ===" 09:18
Woodi eg. if you cache objects you want sometimes to compare them
[ptc] ShimmerFairy: right, ok. Now I just need to work out how to explain that that doesn't come "out of the box"
Woodi [ptc]: there is few way to compare, wrote about one implemented :) 09:19
09:20 bjz left, mr-foobar left
[ptc] Woodi: really? where did you write about it? 09:20
Woodi: do you have a link?
ShimmerFairy [ptc]: I'd guess because it's too hard for a compiler to determine the best way to compare objects (e.g. do you compare private attributes? Are they always to be compared? What if some public attributes don't matter? What if equivalence for the object is more complex?)
09:20 mr-foobar joined
Woodi no, my English is too bad ;) 09:20
[ptc] Woodi: that's ok! :-) 09:21
ShimmerFairy: true. That's probably the best way to explain things
it just needs to be noted in the docs, so that it doesn't trip up others in the future
[ptc] starts hacking the docs again 09:22
09:22 brrt joined 09:23 Begi joined
ShimmerFairy moritz: I think I'd like a hashable concept (though not with that name, I think it gives off the wrong connotation wrt %hashes). I like how smartmatch can be extended by implementing your own ACCEPTS method, and I'd like the same for eqv and other builtin ops. 09:23
[ptc] thanks everyone for the explanations :-)
masak good antenoon, #perl6
yoleaux 06:37Z <smls> masak: You offered to discuss parsing binary blobs, right? I wrote down some thoughts on it here: gist.github.com/smls/bc5d0fb42f199574e339
09:04Z <moritz> masak: the rakudo+nqp 2015.07.1 releases don't work for a star release: "The nqp-m binary is too old" (probably due to the missed version bumb
masak moritz: aww 09:24
that sounds like cause for a 2015.07.2, then.
jdv79 that eqv fallsback to === seems like a bug 09:25
=== is not a "snapshot" cmp 09:26
maybe kick an exception if snapshotting is not possible?
or somehow fix that it is possible in some generic default way
just my naive take 09:27
ShimmerFairy Also, you _can_ define your own === behavior by defining your own WHICH. Seems only fair for eqv to get a method too
jdv79 my main point is that the definition of eqv is not === so why the link?
super confusing for newbies i'd think
moritz jdv79: what would you do instead? die with "Cannot compare two objects of type $typename"? 09:28
ShimmerFairy jdv79: I'm guessing the idea is that === is the most basic and only guaranteed way of checking if two objects of a type are the same. Non-fallback 'eqv' defines a more human comparison of objects (though not as human as ~~) 09:29
jdv79 that's lame but better than a snapshot comp all of the sudden becoming and ident cmp. how does that even make sense?
unless you really want to explain that in here 10 million times;) 09:30
Woodi for me basic equality is if objects *are* the same memory location... 09:31
maybe not only...
probably identity is not the same as equality...
Woodi afks 09:32
09:32 Hor|zon joined
ShimmerFairy I notice that current C<eqv> does effectively a P6 version of C<===>'s NQP-based implementation. It'd be nice if eqv compared on a different method than .WHICH :) 09:33
jdv79 ShimmerFairy: that's not the ways its spec'd 09:34
*way =
ShimmerFairy S03 says on eqv: For object types, compares current contents according to some scheme of canonicalization. 09:36
I'm sure you could argue .WHICH to be some scheme of canonicalization, though in my opinion a poor one.
09:37 Hor|zon left
jdv79 yeah. if this goes out for 6.0 then not sure you could ever change it. that's one concern maybe. 09:38
dalek c: 4c57c15 | paultcochrane++ | lib/Language/operators.pod:
Explain that eqv doesn't work for arbitrary objects

And add a note about how to implement `eqv` if one needs this behaviour. This should address the issue raised in #81.
jdv79 but time constraints i guess
ShimmerFairy moritz: considering S03 calls it "canonical equivalence", how about a multi method CANONICAL() for eqv ? (You could even define the default CANONICAL to just return WHICH, if there's not a better choice of default) 09:39
[ptc] jdv79: I'm guessing perl6/doc #81 is your ticket? 09:40
moritz ShimmerFairy: manybe 09:41
jdv79 i knew the issue sounded familiar
[ptc] feels he understands the issue a bit better now 09:42
jdv79 i was hoping for a lang fix - i think TimToady mentioned it might warrant attention if it was avail
thanks 09:43
09:43 diana_olhovik_ joined
[ptc] jdv79: if the explanation in the docs is still not clear, let me know and I'll try and improve it as best I can 09:43
jdv79 i have to run but i'll look later 09:44
ShimmerFairy yeah, I personally think it'd be helpful for many, if not all of the "complex" comparison ops to call a particular method on the object. That would make it easier to support them, I think. (Esp. if you need to compare private attributes or something similarly better off within the class)
[ptc] jdv79: ok
masak ShimmerFairy: what you're suggesting will late-bind (and pessimise) everything, though. 09:45
jdv79 even a recursive ident cmp would be better to me than just a single top level ===
ShimmerFairy masak: well, ~~ uses ACCEPTS and === uses WHICH (as well as eqv, currently). What would be different about adding CANONICAL? 09:46
moritz masak: fwiw you could still export infix:<eqv> multis for your types
masak smls: gist.github.com/smls/bc5d0fb42f199574e339 looks great. 09:49
vendethiel ShimmerFairy: what's supposed to be the diff. between CANONICAL and WHICH? 09:51
value vs reference equality?
ShimmerFairy vendethiel: A.new(5) eqv A.new(5) should be true, while A.new(5) === A.new(5) typically shouldn't be 09:52
09:52 bjz joined
vendethiel yeah, right. 09:52
seems like "class" vs "data class" in
Scala, say*.
ShimmerFairy m: say [1,2,3] === [1,2,3]; say [1,2,3] eqv [1,2,3] # easy example
camelia rakudo-moar cf30a7: OUTPUT«False␤True␤»
vendethiel would prefer structural equality everywhere anyway :P 09:53
masak won't CANONICAL just end up being Perl 6's hashCode method?
ShimmerFairy What hashCode method?
masak of Java fame. 09:54
hashCode is used to hash things for HashMap entries. 09:55
or hash() from Python, same thing.
ShimmerFairy there's no hashCode in src/core
masak I guess what I mean is, how do you envision implementing a typical CANONICAL method? it will have to do recursive CANONICAL calls on any elements/members the object has, yes?
and you combine all those results in some clever way to make one entire CANONICAL result, yes? 09:56
vendethiel masak: isn't hashCode our WHICH, kinda?
vendethiel thinks he wouldn't like to have both WHICH and CANONICAL, mmh..
masak vendethiel: modulo that we haven't gotten it quite right yet, yes.
ShimmerFairy masak: That's how I imagine it'd be done. I think I just don't like much how 'eqv' will do the wrong thing by default, though admittedly CANONICAL probably wouldn't help that much :) 09:57
masak by the way, I would prepare a 2015.07.2 release, but I'm a little afraid I'll mess it up again :(
09:57 rindolf joined
brrt no, WHICH is equal to id() in python iirc 09:58
masak brrt: WHICH tries to do the job both for reference types and value types presently.
brrt aha
dalek c: a6dbcf1 | paultcochrane++ | lib/Type/Cool.pod:
Mention line-by-line processing in Cool.lines()

This addresses the point made in #83.
ShimmerFairy masak: I just think it's LTA that by default something like CustomInt.new(42) eqv CustomInt.new(42) would return False.
masak ShimmerFairy: I don't think that's unreasonable, though. 09:59
ShimmerFairy: just like you can't auto-deliver deep cloning because you don't know enough about the insides of the user's objects, you also can't auto-deliver eqv.
ShimmerFairy: conceptually, it means that a new class is a black box until its implementor decides otherwise. which feels like a fair/safe deal to me. 10:00
ShimmerFairy m: class A { has $.b }; say A.new(:b(42)) === A.new(:b(42)); say A.new(:b(42)) eqv A.new(:b(42)); say A.new(:b(42)) ~~ A.new(:b(42)); 10:01
camelia rakudo-moar cf30a7: OUTPUT«False␤False␤False␤»
ShimmerFairy masak: I can see that, but it feels like an abnormally low-level thing for Perl 6 to do. Like with the above snippet I ran, I can't help but feel that _at least_ ~~ shouldn't force you to implement something to work right.
10:02 TEttinger left
masak I think defaulting to descending into the user's object isn't a good default. 10:02
ShimmerFairy And I just realized I could probably use your opinion to argue against a default BUILD, if I felt like trying :P 10:03
masak I don't see how my argument applies to that. deep cloning and eqv are *recursive* in nature.
they traverse an object graph, which has all kinds of interesting issues. (cycles being one) 10:04
BUILD happens once and touches one object.
ShimmerFairy What about the attributes that need to be constructed? Which have attributes that need to be constructed? Which... is that not recursive in the same sort of way?
masak ShimmerFairy: answer me this: would your default implementation of eqv for custom objects (a) blow the stack on cycles in the object graph, or (b) be very complicated?
ShimmerFairy I honestly can't answer that, because I have zero clue what (a) is saying. 10:05
masak structurally the choice feels a bit like the choice we're making for `is readonly` for parameters to routines. 10:06
ShimmerFairy masak: I think the current behavior is acceptable, just weirdly low-level for something like P6. It'd be nice if it weren't so, in my opinion.
vendethiel masak:...my opinion on that is that I'd love a "data class" macro that generates WHICH and al
masak that is, we're making the safe/boring choice, because people who care will override it in one of the many possible ways anyway.
and people who don't care will get the safe/boring choice.
vendethiel: sounds fine to me.
masak adds that to his ever-growing list of use cases 10:07
ShimmerFairy I'd argue that it's safe, boring, and highly inaccurate :P
RabidGravy also, I'd suggest the determination whether two objects are actually equivalent things is properly application logic
masak RabidGravy: that, too.
ShimmerFairy: I'm arguing that you want too much magic, in this case, and you seem unaware of the risks of that. 10:08
ShimmerFairy masak: On the one hand I do see how it's too magical for the compiler to do automatically. On the other hand I get the feeling that there will be a lot of basically the same definitions of eqv and the like as it stands, which feels unfortunate. 10:09
vendethiel ShimmerFairy: I think I can write a MOP trait that auto-generates it :)
ShimmerFairy But I don't think you can dispute that it's highly inaccurate for an operator labeled as "canonical equivalence" to think that, by default, canonical equivalence only occurs when you have the exact same object. 10:10
masak I think that's a good, safe default. 10:11
RabidGravy but it's documented like that. "For arbitrary objects this is not possible with the default eqv operator. E.g., eqv will not consider two instances of the same object as being structurally equivalent:" 10:12
masak I don't dispute that it will be wrong in many cases. that's why we allow eqv to be overridden.
ShimmerFairy RabidGravy: in the doc, or the specs?
RabidGravy the doc
ShimmerFairy RabidGravy: that's just describing current behavior. I tend to check the specs for how things should be. (S03 leaves it up to the compiler to decide how "object types" are compared canonically) 10:13
RabidGravy but "structural equivalence" is only something that the implementor of a class can decide, having an overly clever default for arbitrary objects would almost certainly be wrong for a lot of people 10:15
ShimmerFairy masak: An analogy I came up with to describe how I see it would be a default implementation of multiplication as multi sub infix:<*>($a, $b) { 0 } . It's sure true for an infinite number of cases, but it's wrong for an infinite number of cases too. Hence inaccurate.
vendethiel not sure how I can introspect a class' attributes, looking at ClassHOW
I should look at AttributeContainer I guess
masak ShimmerFairy: multiplication also doesn't have to tackle the problem of cycles in the graph and un-canonicalizable objects in the general case. 10:16
ShimmerFairy masak: Like I said, I'm OK with how things stand, I just personally think it's LTA. Not that I think there's a clean or easy solution, I just think it's LTA. :)
masak ShimmerFairy: your analogies keep being compelling only to the extent that they fail to see the depth of the problem.
ShimmerFairy masak: I didn't claim it was an analogy for how I see the problem, just to show in a more simple form why I call it "inaccurate" 10:17
masak ShimmerFairy: I don't like us providing a default that essentially enables very bad failure modes.
ShimmerFairy (Although looking back I don't think I made that clear.)
masak ShimmerFairy: I prefer the implementor of the class to *stop* and *think* in those cases, as they provide their own implementation.
ShimmerFairy masak: I see something like <eqv> as being in the unfortunate middle area where the compiler can't automatically pick a very good default, but a lot of people will write essentially the same code when overriding it. 10:18
masak vendethiel is right -- we should use the MOP and macros to ameliorate that. 10:19
not dangerous defaults.
vendethiel m: multi trait_mod:<is>(Mu:U $type, :$data!){ say $type.attributes; say 1 }; class X is data { has $.a; }; say X.WHAT.perl
camelia rakudo-moar cf30a7: OUTPUT«X␤»
vendethiel I'm not sure why my multi isn't picked up?
masak m: class X is data {} 10:20
camelia ( no output )
moritz m: class X is nosuchtrait { }
camelia ( no output )
masak m: class X is aloogabooga {}
camelia ( no output )
moritz bug!
vendethiel uh.
masak submits it
vendethiel well, seems like I found something.
masak m: class C is aloogabooga {} 10:23
camelia rakudo-moar cf30a7: OUTPUT«5===SORRY!5=== Error while compiling /tmp/UXeCmKzq6m␤'C' cannot inherit from 'aloogabooga' because it is unknown.␤at /tmp/UXeCmKzq6m:1␤»
masak hah!
this time I caught it before hitting "Send" :P
people, do *not* use `X` as the name of an "example class" when testing things on camelia.
use `A` or `C` or some other letter early in the alphabet. 10:24
I'm still going to submit this, because *something* is wrong, clearly.
but it involves using `X`.
vendethiel not sure where "is repr" is defined for nativecall? in rakudo?
m: multi trait_mod:<is>(Mu:U $type, :$data!){ say $type.attributes; say 1 }; class Foo is data { has $.a; }; say Foo.WHAT.perl 10:25
camelia rakudo-moar cf30a7: OUTPUT«5===SORRY!5=== Error while compiling /tmp/swQ_emSAPp␤Method 'attributes' not found for invocant of class 'Foo'␤at /tmp/swQ_emSAPp:1␤»
vendethiel m: multi trait_mod:<is>(Mu:U $type, :$data!){ say $type.^attributes; say 1 }; class Foo is data { has $.a; }; say Foo.WHAT.perl
camelia rakudo-moar cf30a7: OUTPUT«␤1␤Foo␤»
vendethiel I guess that's too early 10:27
10:28 llfourn left
masak it happens right as it parses it, so yes. 10:29
ShimmerFairy I was envisioning some kind of DATAMETHODS() macro that you put in the class, that would supply the appropriate methods, but I'm not sure how possible/sane that is. 10:30
10:31 TEttinger joined
Woodi so object equality is about objec value... if two cars in some point of time drive 100km/h and in that moment we compare them then do they are equal ? 10:38
masak Woodi: it's useful to distinguish "reference comparison" and "value comparison" for exactly that reason. 10:39
Woodi: entity objects tend to very much want to do reference comparison. 10:40
Woodi: (at least until you start doing things like serialization, at which point you probably want to abandon comparing the object system's reference, and instead compare something like an id attribute)
Woodi I hoped we could implement default comparision based on plain attributes like strings and numbers 10:41
10:41 brrt left
masak I see there is a great temptation today to provide this dangerous default :) 10:42
Woodi: and it's surprising you say that after giving your example with two cars.
since two cars with the same driving speed are *not* necessarily the same car. 10:43
Woodi masak: there is moritz++ question :) : do you want to say: Can't compare two object of class "Cls". so looking for default thing for simple situations
masak I don't want us to add a default that only works in "simple situations". 10:44
vendethiel m: class Foo{}; class Foo is bar {};
camelia rakudo-moar cf30a7: OUTPUT«5===SORRY!5=== Error while compiling /tmp/lY0VVmXPMR␤Redeclaration of symbol Foo␤at /tmp/lY0VVmXPMR:1␤------> 3class Foo{}; class Foo is bar7⏏5 {};␤»
vendethiel m: class List is bar {};
camelia rakudo-moar cf30a7: OUTPUT«5===SORRY!5=== Error while compiling /tmp/Ft8f95oemx␤'List' cannot inherit from 'bar' because it is unknown.␤at /tmp/Ft8f95oemx:1␤»
masak that would be like looking at half the problem and saying "eeh, good enough".
vendethiel m: module Foo {}; class Foo is bar {};
camelia rakudo-moar cf30a7: OUTPUT«5===SORRY!5=== Error while compiling /tmp/OqSAraPlMk␤Redeclaration of symbol Foo␤at /tmp/OqSAraPlMk:1␤------> 3module Foo {}; class Foo is bar7⏏5 {};␤» 10:45
vendethiel X is *really* special
.oO( Xspecial )
masak m: say X.HOW
camelia rakudo-moar cf30a7: OUTPUT«Perl6::Metamodel::PackageHOW.new␤»
masak vendethiel: that's why, I think.
vendethiel m: package Foo {}; class Foo is bar {};
camelia rakudo-moar cf30a7: OUTPUT«5===SORRY!5=== Error while compiling /tmp/XzvcBngaPs␤'Foo' cannot inherit from 'bar' because it is unknown.␤at /tmp/XzvcBngaPs:1␤»
vendethiel masak: why isn't this silent then?
masak m: package Foo::Yay {}; class Foo is bar {} 10:46
camelia rakudo-moar cf30a7: OUTPUT«5===SORRY!5=== Error while compiling /tmp/4818O_LhZd␤'Foo' cannot inherit from 'bar' because it is unknown.␤at /tmp/4818O_LhZd:1␤»
masak vendethiel: no idea.
jnthn X is the package where all the exceptions live. 10:54
masak jnthn: yes, but people seem to forget that when testing something on camelia. 10:55
jnthn masak: They seem to be quickly reminded :P
vendethiel I sure as hell always do. :)
jnthn masak: I actually patched the compiler to only look for errors it throws in SETTING:: a while ago 10:56
masak jnthn: not always. there are a couple of RT tickets filed because people didn't realize.
jnthn masak: They realized something was wrong enough to file an RT :P
masak troo 10:57
jnthn I'm not sure what we can easily do about it.
Getting the compiler to only look for compile-time exceptions in the setting rather than anything the user has declared got rid of one set of ways to get confused.
m: class X { }; omg wtf bbq
camelia rakudo-moar cf30a7: OUTPUT«5===SORRY!5=== Error while compiling /tmp/er6FsXMftg␤Undeclared routines:␤ bbq used at line 1␤ omg used at line 1␤ wtf used at line 1␤␤»
jnthn star: class X { }; omg wtf bbq 10:58
camelia star-m 2015.03: OUTPUT«===SORRY!===␤Error while compiling, type X::Undeclared::Symbols␤ post_types: (unstringifiable object)␤ unk_types: (unstringifiable object)␤ unk_routines: (unstringifiable object)␤ routine_suggestion: (unstringifiable object)␤ type_suggestion: …»
jnthn Which fixed things like htat
m: class X { }; omg wtf!!!
camelia rakudo-moar cf30a7: OUTPUT«5===SORRY!5=== Error while compiling /tmp/QZ8ZMX_K9X␤Bogus postfix␤at /tmp/QZ8ZMX_K9X:1␤------> 3class X { }; omg wtf7⏏5!!!␤ expecting any of:␤ infix␤ infix stopper␤ postfix␤ statement end␤ …»
jnthn star: class X { }; omg wtf!!!
camelia star-m 2015.03: OUTPUT«===SORRY!===␤Error while compiling, type X::Syntax::Confused␤ reason: Confused␤ at line 2, near "!!!"␤Error while constructing error object:Could not locate compile-time value for symbol Syntax::Confused␤»
jnthn That's a better example.
masak heh. "confused. reason: confused" :P
jnthn It'd be good to see if any other the other RTs about it are covered by the fix I already did.
On the eqv default thing and avoiding a dangerous-but-tempting result, I agree with masak++ 11:02
uh, s/result/default/
ShimmerFairy m: package S { }; class S is foobar { } 11:03
camelia rakudo-moar cf30a7: OUTPUT«5===SORRY!5=== Error while compiling /tmp/hTH67oTftq␤'S' cannot inherit from 'foobar' because it is unknown.␤at /tmp/hTH67oTftq:1␤»
ShimmerFairy ^ only doesn't error with the "X" package, AFAICT
11:09 rindolf left, darutoko left, darutoko joined
dalek c: 12b62f0 | paultcochrane++ | lib/Language/functions.pod:
Indent subroutine example code consistently with rest of text
 11:11
c: a34c050 | paultcochrane++ | lib/ (2 files):
Indent IO.lines examples consistently
11:11 TEttinger left, jack_rabbit left 11:13 lizmat left 11:15 yakudza left, rindolf joined 11:16 itz joined 11:17 llfourn joined 11:19 itz_ left
dalek c: a404865 | paultcochrane++ | lib/Type/IO/ArgFiles.pm:
Add docs for IO::ArgFiles
 11:20
11:28 mr-foobar left 11:29 mr-foobar joined 11:33 Hor|zon joined 11:37 Hor|zon left 11:44 araujo left 11:46 Alina-malina left 12:00 Averna left 12:14 bin_005 left 12:17 AlexDaniel joined 12:21 colomon left 12:29 diana_olhovik_ left 12:34 Hor|zon joined 12:38 Hor|zon left
RabidGravy If I wrap a C++ class to make a C interface so I can use it with NativeCall and one of the underlying methods throws an exception, that exception can't be cauught in perl space can it? 12:39
jnthn RabidGravy: No, that's required quite some magic. Maybe FROGGS++'s nativecall C++ work will be able to do that at some point 12:43
But it's probably highly non-trivial
RabidGravy yeah that's what I thought, liberally sprinkly try catch all over the place then 12:52
13:10 Alina-malina joined 13:21 rindolf left 13:23 rindolf joined
moritz masak: are you doing the .2 releases, or should I? 13:38
13:47 lizmat joined 13:50 rindolf left 13:52 llfourn left 13:53 FROGGS_ joined 13:57 FROGGS left
dalek p: a4fcfa7 | moritz++ | VERSION:
Version 2015.07.2

this time with the correct VERSION number		 14:00
14:08 rindolf joined 14:10 Mhoram left 14:11 Mhoram joined 14:14 lucasb joined 14:16 llfourn joined 14:18 vendethiel left 14:25 JimmyZ_ joined 14:26 cognominal joined 14:27 brrt joined 14:35 Hor|zon joined
dalek kudo/nom: 42a3314 | moritz++ | / (2 files):
Bump NQP revision and VERSION to 2015.07.2
 14:37
kudo/nom: 7664529 | moritz++ | tools/build/Makefile-Moar.in:
Build: add sanity check that version in VERSION and tarball version are the same
kudo/nom: 6bbb56f | moritz++ | docs/release_guide.pod:
Note 2014.07.2 release in release guide
moritz running final (I hope) spectest with 2015.07.2 14:38
oh, and typo in previous commit message :(
14:39 Hor|zon left
dalek p: 16bee5d | moritz++ | tools/build/Makefile-Moar.in:
Build: catch differing versions between tarball and VERSION file
 14:45
lucasb it will ship with the '~~' change? 14:46
14:53 dolmen joined 14:54 skids joined 15:01 brrt left 15:04 mr-foobar left 15:13 [Tux] left 15:25 brrt joined 15:26 [Tux] joined 15:35 dolmen left
lizmat lucasb: it looks like it 15:37
brrt \o
timotimo ugexe: an animated gif wouldn't have gone amiss for your progress bar example 15:39
alternatively, re-implement the animation in javascript 15:40
:P
lizmat timotimo brrt o/
brrt hmm, i'm puzzling on how to make the final ruleset table, and how to represent it in memory 15:42
option one: nested binary trees; option two, hash tables 15:43
lizmat moritz: did you upload 2015.07.2 already ?
brrt oh, and it has to work as a linear compiled array, please :-)
lizmat brrt: I'm afraid nothing comes to mind at this early hour 15:44
brrt hmm
lizmat (08:45am)
brrt that's ok, of course:-)
17:45 :-P
moritz lizmat: only nqp 15:45
lizmat ok, so we *could* revert the ~~ chaining change for the 2015.07.2 still ?
15:46 Averna joined
moritz lizmat: yes; should we? 15:46
ugexe timotimo: is there a suggested tool for creating such a gif?
lizmat as that seems to have some ecosystem fallout
moritz ok
should I revert it on nom, or in a release branch?
lizmat not sure at this messy point :-) 15:47
moritz I tend towards a release branch
brrt probably release branch
lizmat there you go!
:-)
brrt i mean, the change was probably introduced with Good Reason
lizmat yeah, but we want an easy rakudo star release :-)
moritz thus, release branch 15:48
dalek kudo/release/2015.07.2: 06fc07e | moritz++ | src/Perl6/Grammar.nqp:
Revert "move ~~ from chaining to structural"

This reverts commit 81ad2c0db9910fe7093c1db1d0e25e4710f3e67e.
This change was introduced too short before a release to deal with the ecosystem fallout in time for R*
moritz t/spec/S12-attributes/mutators.t aborts here
brrt is not currently on a sufficiently fast computer to test it 15:49
hahainternet has a ludicrously fast computer should you require it
lizmat moritz: it has skips in it, did you run the .t with perl6 ?
moritz lizmat: I just ran 'make spectest' and later 'make t/spec/S12-attributes/mutators.t'
lizmat moritz: make runs fine for me 15:50
(on OS X)
15:51 Hor|zon joined
moritz perlpunks.de/paste/show/55b3b06c.54a2.91 15:51
oh wait, I have local modifications in that file
never mind me.
brrt be back much later 15:52
hahainternet brrt: what do you need? 15:53
oh ok
brrt moritz++ needs a spectest run ;-)
brrt needs time to think
hahainternet well i have some overclocked i7
15:53 brrt left
hahainternet so shout me if you need it 15:53
moritz 2015.07.2 uploaded
hahainternet: I'm fine, thanks
release tagged and pushed 15:55
15:55 Hor|zon left
lizmat moritz++ 15:59
16:00 grondilu joined
grondilu can someone give me an ELI5 about this: github.com/azawawi/perl6-atom-electron 16:00
16:01 xinming left
timotimo ugexe: no clue :\ 16:01
i bet vlc can do it :P
tadzik I don't know, but it looks awesome 16:02
grondilu guesses he should check electron.atom.io/ out
16:11 lolisa joined
lizmat decommute, see you on the flipside& 16:12
16:12 lizmat left 16:23 llfourn left
moritz just tried to do a star, and Bailador fails (wants Template::Mustach), DBIish fails a test (30-Pg.t), and doc needs File::Temp in a test 16:27
16:34 jkva joined
jkva Good localtime. I'm experiencing a build fail during installation of Task::Star. Is this an appropriate channel to ask for assistance? 16:35
ugexe added a gif, good suggestion 16:37
16:37 skids left
RabidGravy jkva, could be. Ask the question and we'll let you know ;-) 16:40
16:41 Begi left
jkva RabidGravy: Gotcha :) I'm running Debian Squeeze, just finished installing rakudobrew, moarvm and panda. When running "panda install Task::Star" as per the directions, I eventually get: 16:41
the following build error - pastie.org/10311742 16:42
TimToady ShimmerFairy: the problem with defining a CANONICAL is that eqv really wants to compare subparts lazily, so that if the first subpart doesn't match, you never have to canonicalize the rest of the object
ShimmerFairy TimToady: interesting. How would laziness be ensured by an object who defines eqv as basically $a.A eqv $b.A && $a.B eqv $b.B && ... (assuming a non-shortcircuiting version of the example I typed) 16:46
TimToady I think a HOP solution would be better, where the class can return a function that can do the comparison
ShimmerFairy Unless each attribute comparison isn't considered a subpart, in which case I'm not quite clear on what "subpart" means :)
HOP = Higher Order Perl [5] ? 16:47
TimToady Programming
RabidGravy jkva, I believe that is caused by a very rcent change in rakudo. It will be fixed in the modules 16:48
TimToady attribute by attribute, yes, but that could also be broken down if there were, say, a hash value on an attribute to do a fast rejection before doing the harder slow path 16:49
jkva RabidGravy: Too bad. If I want to play with Perl6, do I have any other means at this point? Use an older version?
16:49 AlexDaniel left
TimToady there's a .2 release that addresses that 16:50
though it's in a branch 16:51
ShimmerFairy TimToady: when you say "a function that can do the comparison", are you referring to an ACCEPTS-like function? And would that solution require changes to the current core, or could it be done entirely within user code?
RabidGravy jkva, github.com/rakudo/rakudo/tree/release/2015.07.2 specifically 16:52
jkva Ah, excellent. Thank you very much, I will try my luck with that one. 16:53
TimToady at the moment you'd have to call that from your own multi, so there's not much point, but if there were a way that built-in eqv could defer to your EQV function, then you wouldn't have to write a multi
I think the built-in eqv should guarantee that the type is the same before delegating, so we don't give the class the option of defining a non-symmetric eqv 16:56
RabidGravy sounds perfect to me
ShimmerFairy TimToady: so basically "just like ~~ and ACCEPTS", instead of my less-sensible idea of "just like === WHICH", right? :)
TimToady: I recall a syntax for ensuring that, I used it in my old S32::Stringy gist...
TimToady well, though ACCEPTS is far from symmetrical
ShimmerFairy Sure, I just mean "calling a function with arguments" vs. "calling a function that returns a string of info" 16:58
TimToady and ACCEPTS is a direct action, whereas I was thinking more of a method to be called at compile time that returns a function that performs the action so we get compile-time dispatch
nine my $foo = sub { ... }; gives me an anonymous sub. How can I create a named sub? 16:59
TimToady where do you want to put the name?
ShimmerFairy m: sub foo(Any ::T $a, T $b) { $a ~~ $b }; say foo(1, 1); say foo(1, "1"); # how you would ensure the same type in a signature, for reference 17:00
camelia rakudo-moar 6bbb56: OUTPUT«True␤Type check failed in binding $b; expected 'Int' but got 'Str'␤ in sub foo at /tmp/KVrwhme2xn:1␤ in block <unit> at /tmp/KVrwhme2xn:1␤␤»
nine TimToady: I'm binding the thus created sub using ::($module).WHO{"&$name"} := sub ... I'd like ::($module).new.^methods to give better answers than a long list of <anon> <anon> 17:01
TimToady you don't want to use ~~ for eqv
jnthn nine: .set_name on the sub
TimToady but ^methods won't get the name 17:02
ShimmerFairy TimToady: sure, that was just a silly example sub, the point was demonstrating ::T (admittedly I should've used an irrelevant-to-the-discussion op there, instead of ~~)
TimToady why I asked where the name is supposed to go
17:02 llfourn joined, telex left
TimToady ShimmerFairy: you'd also have to catch the bind error and return false instead 17:02
nine Unfortunately TimToady seems to be right. Still getting <anon> 17:03
TimToady nine: note that P6 does not store method names in packages anymore, is what we're trying to say
ShimmerFairy You could do that with a multi sub with arguments (Any $a, Any $b), right?
TimToady so you'd have to do an add_method somewhere to put it into the metaobject
jnthn walk &
17:04 vendethiel joined
ShimmerFairy m: multi sub foo(Any ::T $a, T $b) { True }; multi sub foo(Any $a, Any $b) { False }; say foo(1, 1); say foo(1, "1"); 17:04
camelia rakudo-moar 6bbb56: OUTPUT«True␤False␤»
17:04 telex joined
ShimmerFairy Unless that's suboptimal in some way for eqv , that looks like it would work (it _might_ be useful to make eqv do that now anyway, since I don't think it does) 17:05
17:06 llfourn left
TimToady seems to me that it would be better to factor out the type test to before we ever dispatch to any submatchers, rather than scattering the test to every type, and allowing the possibility that someone could write one without the type test accidentally or on purpose 17:07
that seems like a worthwhile invariant to enforce
RabidGravy yes 17:08
jkva Unfortunately I am still experiencing the same error when installing Task::Star - I checked out release 2015.07.02 and built rakudo, moarvm and panda manually. "This is perl6 version 2015.07.1-3-g6bbb56f built on MoarVM version 2015.07" 17:24
Although the perl6 version does not seem to match the release number. Perhaps I made a mistake somewhere? 17:25
TimToady it's in a branch, did you checkout? 17:26
jkva TimToady: Most likely I errored there, I'll retry. 17:27
17:31 lolisa left
TimToady masak: we have to solve the cycles problem for serialization anyway, so the correct default is probably just to call a cycle-aware serializer on both objects and compare the resulting buffers; a default clone should probably be using the same default definition via serialize/deserialize 17:32
then eqv optimizations are just short-circuiting "would these objects serialize the same if we tried"
and deepclone optimization is bypassing the actual construction/deconstruction of the serialized form 17:33
but the definition of what is serializable stays the same, abstracted to one spot 17:34
I'd rather have a slow correct eqv and deepclone that for it to default to the wrong sematics
*than
then "can I call eqv/deepclone?" semantically reduces to "are these serializable?" 17:35
17:40 rangerprice joined, konsolebox left
TimToady and standard policies for detecting cycles also need to be available to people writing their own eqv, so that they can behave as if a massive serialization were being done by the top-level eqv, without actually doing the massive serialzation 17:48
which implies we know how to detect a top-level eqv and distinguish it from eqv helpers
jkva Got past the error, excellent. I made a mistake checking out the release. Thanks! 17:49
TimToady yer welkum 17:50
17:52 Hor|zon joined 17:56 Hor|zon left
jkva Wth? DBDish expects me to have mysql running for its test? 17:56
Shouldn't that be mocked? 17:58
Ah, I see on github it's on the agenda. Nvm. 18:01
x§ 18:05
18:12 zakharyas joined 18:24 zakharyas left
moritz jkva: DBIish comes with the sqlite, mysql and postgres backends 18:29
jkva: and it tries to run all of its tests, but skip those tests that it can't run due to missing libs 18:30
18:32 apejens left 18:33 apejens joined 18:40 jkva left 18:42 zakharyas joined 18:45 cognominal left, cognominal joined
lucasb m: BEGIN $_ = 1 18:55
camelia ( no output )
lucasb m: BEGIN { $_ = 1 } 18:56
camelia rakudo-moar 6bbb56: OUTPUT«5===SORRY!5=== Error while compiling /tmp/tpGfrQflqt␤An exception occurred while evaluating a BEGIN␤at /tmp/tpGfrQflqt:1␤Exception details:␤ 5===SORRY!5=== Error while compiling ␤ Cannot assign to an immutable value␤ at :␤»
dalek Iish: cc432b0 | moritz++ | lib/DBDish/Pg.pm6:
Fix test failure in t/30-Pg.t

I think this is a workaround for a recently-ish rakudobug
lucasb m: INIT $_ := 1; .say 18:59
camelia rakudo-moar 6bbb56: OUTPUT«1␤»
lucasb m: INIT { $_ := 1 }; .say
camelia rakudo-moar 6bbb56: OUTPUT«(Any)␤»
moritz a block has its own $_ 19:01
nine m: my $name = "Foo"; my $class := Metamodel::ClassHOW.new_type(name => $name); ::($name) := $class; 19:02
camelia rakudo-moar 6bbb56: OUTPUT«5===SORRY!5=== Error while compiling /tmp/4ZSHivYBNG␤Cannot use bind operator with this left-hand side␤at /tmp/4ZSHivYBNG:1␤------> 3type(name => $name); ::($name) := $class7⏏5;␤»
nine That's an LTA error message. It explains neither why I cannot use the bind operator, nor what an alternative could be.
lucasb moritz: So it would never work? Well, that explains it. thanks.
TimToady m: INIT { OUTER::<$_> := 1 }; .say 19:04
camelia rakudo-moar 6bbb56: OUTPUT«1␤» 19:05
TimToady m: INIT { UNIT::<$_> := 1 }; .say
camelia rakudo-moar 6bbb56: OUTPUT«1␤»
19:05 AlexDaniel joined
TimToady m: BEGIN { OUTER::<$_> = 1 }; .say 19:06
moritz nine: it's hard make that error better
camelia rakudo-moar 6bbb56: OUTPUT«5===SORRY!5=== Error while compiling /tmp/r34LRgQ9JY␤An exception occurred while evaluating a BEGIN␤at /tmp/r34LRgQ9JY:1␤Exception details:␤ Cannot modify an immutable Mu␤ in block at /tmp/r34LRgQ9JY:1␤␤»
nine So what's wrong with this? ::($name) := $class Or rather, what would be the real way to make a class accessible through its name?
moritz nine: you cannot add stuff to lexpads at run time
nine moritz: I can if I use EVAL, so I figured that there should be some MOP way, too? 19:07
moritz nine: with EVAL you can only add it to the lexpad inside the EVAL
TimToady you can only add it to EVAL's lexical scope, not its OUTER
moritz nine: if you want to install it as an our-symbol, you can assing to GLOBAL::($name), I think 19:08
or was it GLOBAL::{$name}?
tadzik: currently Bailador depends on both Template::Mustache and Template::Mojo, forcing Star to include both. Is that intentional? 19:09
nine Indeed I use GLOBAL:: in the EVAL: github.com/niner/Inline-Perl5/blob...5.pm6#L698 19:10
m: my $name = "Foo"; my $class := Metamodel::ClassHOW.new_type(name => $name); GLOBAL::($name) := $class;
camelia rakudo-moar 6bbb56: OUTPUT«5===SORRY!5=== Error while compiling /tmp/0K_Y18FfpL␤Cannot use bind operator with this left-hand side␤at /tmp/0K_Y18FfpL:1␤------> 3ame => $name); GLOBAL::($name) := $class7⏏5;␤»
nine Unfortunately still the same error
moritz m: GLOBAL::{'Foo'} := class { };
camelia ( no output )
moritz nine: works with {}
TimToady ::() is just symbolic lookup, ::{} is actually using the Stash as a hash, so it can give you a bindable 19:12
dalek ar: 814f76b | moritz++ | / (3 files):
Initial preparations for 2015.07
nine moritz: thanks! One step further.
m: BEGIN { my $name = "Foo::Bar"; my $class := Metamodel::ClassHOW.new_type(name => $name); $class.^compose; GLOBAL::{$name} := $class; }; Foo::Bar.new;
camelia rakudo-moar 6bbb56: OUTPUT«Could not find symbol '&Bar'␤ in block <unit> at /tmp/Kol65Ba5qQ:1␤␤Actually thrown at:␤ in block <unit> at /tmp/Kol65Ba5qQ:1␤␤»
dalek ar: 098e45e | moritz++ | modules/ (10 files):
Update submodules
moritz nine: probably need to split on :: yourself, and plot your way through GLOBAL::{$first}::{$second} 19:13
tadzik moritz: hmm, possibly. I guess it could be made an optional dependency 19:14
nine moritz: yep. I seem to remember having written such code, came across some problem and frustratedly giving up and use EVAL instead 19:15
dalek ar: a383d29 | moritz++ | modules/ (3 files):
Remove deleted subodules more thoroughly
 19:19
RabidGravy I'm quite liking subodules 19:22
moritz RabidGravy: I hate them 19:23
RabidGravy :)
moritz the submodules are constantly in "detached HEAD" state, which makes working in them a PITA
they have multiple redundant locations for meta data; changing only one of them leaves them in inconsistent state 19:24
'git archive', 'git grep' et al ignore submodules
the list goes on.
flussence moritz: you're not the only one; I've noticed most projects that use submodules wrap them in a ton of scripting so nobody has to touch git for them directly... 19:28
moritz ok, star-m now passes all tests
flussence: I know; #git recommends not using them either, mostly 19:29
19:29 ggoebel joined
flussence I'm hoping somewhere down the line an alternative idea'll pop up that people agree on being good enough, sort of like Moose did for p5 19:32
19:32 darutoko left
flussence git's really-simple-core design seems like it'll leave space for that to happen 19:32
moritz well, subtree exists, for cases where size isn't a problem 19:33
19:34 jkva joined
jkva moritz: For me the mysql test failed, and stopped Task::Star from going on to the next module. I resorted to using -notests, but would rather not have. 19:35
moritz: ah, I understand. I have libraries installed, but not MySQL itself. So this is an error on my part. Apologies, I was mistaken.
moritz jkva: no, souldn't fail in that case either
jkva moritz: Would you like me to file an issue? 19:36
moritz jkva: yes please
jkva moritz: will do.
moritz jkva++
19:38 ggoebel left 19:48 yqt joined
jkva moritz: I've opened github.com/perl6/DBIish/issues/14 - keep in mind I am really new to Perl6, the problem might very well be on my end. Please let me know if you need more info. 19:51
19:52 Hor|zon joined
moritz jkva: ah, it's actually t/30-Pg.t that fails 19:53
jkva: the mysql tests spew out error messages, but don't fail
jkva d'oh
moritz jkva: and the failure in 30-Pg.t should be fixed by github.com/perl6/DBIish/commit/cc4...186fe3086c (from an hour ago, roughly) 19:54
jkva Ha, no way
Awesome
Right, the mysql errors were a red herring. Thanks! 19:55
dalek Iish: 7cf60c4 | moritz++ | lib/DBDish/Pg.pm6:
Fix a build failure with rakudo-2015.07, jkva++

closes #14
jkva Whelp, that one was quickly sorted :P
moritz jkva: but the end of your paste also include an error that I hadn't seen before, which the commit above should have fixed 19:56
jkva moritz: glad to hear - I figured it couldn't hurt
19:57 Hor|zon left 20:11 bin_005 joined
moritz jkva: anyway, hope it's all good now 20:11
jkva moritz: I expect it to be - I'm going to make a fresh install of Task::Star later. Do you know of a preferred way of reinstalling already installed modules with panda? 20:13
moritz jkva: no idea 20:15
jkva moritz: No probs - I'll figure things out. Thanks for the quick fixes, moritz++
RabidGravy jkva, if you do "panda install <module>" it usually just reinstalls it 20:21
jkva RabidGravy: thanks, worth a shot :)
RabidGravy: you are correct. moritz: all good now! 20:24
RabidGravy for reference if you want to reinstall *all* modules, reinstalling panda will cause that 20:25
20:25 zakharyas left
moritz \o/ 20:25
20:26 rindolf left
jkva RabidGravy: Yeah, although I'd rather avoid that :) 20:26
20:40 lucasb left 20:55 jkva_ joined 20:57 jkva left 21:00 kaare__ left 21:06 rangerprice left 21:08 jkva_ left 21:16 yqt left 21:19 yqt joined 21:33 Begi joined 21:38 mdinger joined 21:48 jkva joined 21:53 Hor|zon joined 21:58 jkva left, Hor|zon left, jkva joined
dalek ast: f4caecb | PerlJam++ | S32-array/splice.t:
Add test for RT#119913
 22:04
synbot6 Link: rt.perl.org/rt3/Public/Bug/Display...?id=119913
RabidGravy (more tests)++ 22:08
22:21 perturbation joined 22:25 Begi left
mdinger Is there a reason lookaround uses really weird ordering? For example: `rx{ <?after foo> bar <?before baz> }` has `after` before `bar` and `before` after `baz`...reading it is pretty awkward. Shouldn't it be backwards? 22:32
vendethiel m: say 'foobarbaz' ~~ /<?after foo>bar<?before baz>/; # mdinger 22:34
camelia rakudo-moar 6bbb56: OUTPUT«｢bar｣␤»
vendethiel mdinger: basically, the first "after" says "what i just saw contained foo". then, the before says "what i WILL see will be baz" 22:35
mdinger vendethiel: okay. It does make sense and I wasn't questioning if it worked. 22:38
22:38 jkva left
mdinger vendethiel: It just seems weird. Is it written from the perspective of the parser? 22:39
vendethiel mdinger: no, it's written in the order I expect. It's not "surrounding" bar. 22:41
"my content is right AFTER what I just saw" (after). "you see bar" (bar) "my content must be BEFORE what I will see" (<before>)
22:49 TEttinger joined
mdinger vendethiel: okay. This'll take some work I guess. I really really want to read it like this I guess: `/<?before foo>bar<?after baz>/` foo is before bar. baz is after bar. 22:50
I can read it as you showed I just have to be really careful...maybe it'll come with time 22:52
dalek ast: e8821d1 | PerlJam++ | S16-filehandles/io.t:
Add some tests for RT#113100
synbot6 Link: rt.perl.org/rt3/Public/Bug/Display...?id=113100
dalek c: 6b4f45c | mdinger++ | lib/Language/regexes.pod:
Document lookahead assertions
 23:05
c: a186323 | RabidGravy++ | lib/Language/regexes.pod:
Merge pull request #106 from mdinger/lookahead

Document lookahead assertions
RabidGravy there you go :) 23:06
mdinger thanks guys 23:08
23:08 mdinger left, mr-foobar joined
dalek ast: 8c8efe8 | PerlJam++ | S02-types/bool.t:
Add a couple of todo tests for RT#72580
 23:12
synbot6 Link: rt.perl.org/rt3/Public/Bug/Display...l?id=72580
23:13 [Tux] left 23:15 JimmyZ_ left 23:33 BenGoldberg joined 23:35 TimToady left 23:37 TimToady joined
RabidGravy nighty night 23:38
timotimo nn
23:42 RabidGravy left 23:54 Hor|zon joined 23:59 Hor|zon left

Powered by Cro and the Raku® Programming Language.

Please report any issues / comments / feature requests as an issue on App::Raku::Log.

Thank you!